Generate SEO-friendly URL slugs from titles in any language — powered by DeepSeek V4.1 Flash with thinking disabled for fast, direct output.
Why AI instead of plain rules: a rule-based slugger can only transliterate characters (在家煮咖啡 → zai-jia-zhu-ka-fei). This tool translates the meaning into concise English keywords (在家煮咖啡 → brew-coffee-at-home), drops stopwords like "the/of/and", and puts the highest-value search keywords first — the pattern recommended by Google's URL structure guidelines (hyphens, lowercase, short).
How it works:
- Your title goes to DeepSeek V4.1 Flash (
deepseek-flash) in non-thinking mode (thinking: {"type": "disabled"}, the official parameter) — no reasoning tokens, direct answer, low latency and cost. When only an OpenRouter key is configured, the same generation runs as deepseek/deepseek-flash-latest with reasoning: {"enabled": false}.
- Every slug from the model is re-checked by a deterministic sanitizer: NFKD fold to ASCII, lowercase, single separators, whole-word truncation inside your character budget. A malformed model answer can never leak into your URL.
- You get one primary slug plus up to 4 alternatives, extracted keywords, detected language, and a one-line rationale.
- Offline fallback: if no API key is configured or the model call fails, the tool degrades gracefully to a rule engine (CJK → tone-less Pinyin via
pinyin-pro, diacritics folded) and clearly marks the result as degraded. You can also force it with Rules only (offline).
Slug conventions applied: lowercase ASCII, hyphen (or underscore) word separators, 3–5 words by default, ≤60 characters by default, stopwords removed, most important keyword first.